Low-Frequency Electromagnetic Waves on a

Twisted-Pair Transmission Line

Kirk T. McDonald
Joseph Henry Laboratories, Princeton University, Princeton, NJ 08544
(December 24, 2008)
1 Problem
Discuss the electromagnetic waves that can propagate in the space around a transmission
line whose form is a double helix of radius a and longitudinal period p a. The pitch angle
of the helical windings with respect to the transverse planes is given by
cot = k
p
a =
2a
p
. (1)
The angle of the windings with respect to the axis of the line is then = /2 , i.e.,
tan = k
p
a. (2)
Such lines are extensively used for telephone communication at low frequencies for which
ka, kp 1, where k = 2/ = /v is the wave number at angular frequency , and v is
the wave velocity. For the case that ka, kp 1 the waves can be thought of following the
helical conductors such that the group velocity along the axis of the helix is
v
g,z
c cos . (3)
Show that even at low frequencies eq. (3) is a reasonable approximation when a p, but
when a p (a gentle twist) then v
g,z
c

cos .
1
2 Solution
Despite the common use of twisted-pair transmission lines, this problem seems little dis-
cussed in the literature. In the case of two-dimensional conductors there exist transverse
electromagnetic (TEM) waves of the form e
i(kzt)
times the (transverse) static electric and
magnetic eld patterns. However, TEM waves will not propagate along a twisted pair of
wires, whose structure is three-dimensional.
Waves on a single helical conductor have been discussed in the context of traveling-wave
ampliers in the sheath approximation [1, 2], where only the part of the waves that are
independent of azimuth are analyzed. A fairly general discussions of waves on twisted-pair
conductors for ka kp 1 has been given in [3], again in the context of traveling-wave
ampliers.
1,2
Here, we emphasize the low-frequency behavior, when ka, kp 1.
2.1 General Form of the Fields in Cylindrical Coordinates
We use a cylindrical coordinate system (r, , z) whose axis is that of the transmission line.
We ignore the insulation typically found on the wires of a twisted-pair line, and assume that
the space outside the wires is vacuum.
The electromagnetic elds Eand Bwith time dependence e
it
satisfy the vector Helmholtz
equation,
(
2
+ k
2
f
)E, B = 0, (4)
outside the wires, where
k
f
=

c
=
2

f
. (5)
However, in cylindrical coordinates only their z-components satisfy the scalar Helmholtz
equation,
(
2
+ k
2
f
)E
z
, B
z
= 0. (6)
We look for wavefunctions for E
z
and B
z
that propagate in the z-direction with the form
f
m
(r)e
im
e
i(kmzt)
, (7)
where m is an integer. The (right-handed) helical conductor rotates by = k
p
z = 2z/p as
z increases, so we expect the wavefunction (7) to include this symmetry via a phase factor
e
im(kpz)
such that the waveform rotates as it advances. The z-dependent part of this
phase contributes to the wave number k
m
, which takes the form
3
k
m
= k
0
() + mk
p
. (8)
1
See [4] for the case of cross-wound helices.
2
The magnetic elds of twisted pairs have been discussed in [5, 6, 7, 8]. Twisted-pair structures with
large currents are used as undulators to generate energetic photon beams at particle accelerators (see, for
example, [9]).
3
The present case contrasts with that of so-called Bessel beams of order m(see, for example, the Appendix
of [10]) where the drive currents are limited to a small region in z, rather than being periodic in z, such that
k
m
= k
0
for any index m.
2
We are mainly interested in waves that propagate in the +z direction, for which the index
m must be non-negative at low frequencies where 0 < k
0
k
p
.
4
The phase
m
of the wave function (7) is
m
= k
(m)
x t = k
m
z m t, where
the wave vector k
(m)
is given by
k
(m)
=
m
= k
m
z
m
r

. (9)
The phase velocity v
p,m
of a partial wave of index m is
v
p,m
=

k
(m)

k
(m)
=
ck
f
k
2
m
+ m
2
/r
2
_
k
m
z
m
r

_
. (10)
We expect that k
0
<

k
f
( k
p
) so that v
p,0
<

c z, but for nonzero index m we have that

k
m
mk
p
, and hence
v
p,m

ck
f
r
m[1 + (k
p
r)
2
]
_
k
p
r z

_
, (11)
which is small compared to c at any value of r. The wave vector k
(m)
(and the phase velocity
v
p,m
) make angle
k
to the z-axis given by
tan
k
=
1
k
p
r
(12)
for any nonzero index m. Note that at r = a the wave vector is at right angles to the
direction of the helical windings, for which tan = k
p
a.
The group velocity of a partial wave is
5
v
g,m
=
k
(m
)
=

k
(m)
, (13)
whose only nonzero component is
v
g,m,z
=
d
dk
(m)
z
=
d
dk
m

1
dk
m
/d
=
1
dk
0
/d
= v
g,0,z
v
g,z
, (14)
independent of index m. We expect that v
g,z
<

c in the low-frequency limit.

Using eqs. (7)-(8) in the Helmholtz equation (6), we see that the radial function f
m
obeys
the Bessel equation
1
r
d
dr
_
r
df
m
dr
_

_
k
2
m
k
2
f
+
m
2
r
2
_
f = 0, (15)
where |k
m
| k
0
> k
f
. The solutions to eq. (15) should remain nite at r = 0 and , so for
r < a we use the modied Bessel function I
m
(k

m
r), and for r > a we use K
m
(k

m
r), where
k

m
=
_
k
2
m
k
2
f
. (16)
4
Waves with index m negative (both for single helix and double-helix congurations) have their phase
and group velocities in opposite directions. An application of such waves is the backward wave oscillator. See,
for example, [11].
5
See, for example, sec. 2.1 of [12].
3
That is, the longitudinal components of the electric and magnetic elds outside the wires
have the forms
E
z
(r < a) =

m
E
m
I
m
(k

m
r)
I
m
(k

m
a)
e
im
e
i(kmzt)
, E
z
(r > a) =

m
E
m
K
m
(k

m
r)
K
m
(k

m
a)
e
im
e
i(kmzt)
,
(17)
B
z
(r < a) =

m
B
m
I
m
(k

m
r)
I

m
(k

m
a)
e
im
e
i(kmzt)
, B
z
(r > a) =

m
B
m
K
m
(k

m
r)
K

m
(k

m
a)
e
im
e
i(kmzt)
,
(18)
where B
m
and E
m
are constants to be determined, and I

m
(k

m
a) = dI
m
(k

m
a)/dr. In eq. (17)
we have noted that the Maxwell equation E = ik
f
B (in Gaussian units) implies that E
z
(and E

) is continuous across the surface r = a. We verify later that the normalization of

coecients B
m
to I

m
(k

m
a) and K

m
(k

m
a) insures continuity of the magnetic eld component
B
r
across this surface, as required by the Maxwell equation B = 0.
The waves are driven by the current density J in the twisted pair, which we can write as
J(x, t) = J(, z, t)(r a)(sin

+ cos z), (19)
which points along the local direction of the twisted-pair conductors, and is conned to a
thin cylinder of radius a. The wavefunction J(, z, t) must have the same dependence on ,
z and t as eqs. (17)-(18), namely
J(, z, t) =

m
J
m
e
im
e
i(kmzt)
, (20)
assuming that the current only ows in the direction of the helical windings.
For a twisted pair, the current at xed z and azimuth + is opposite to that at azimuth
, which implies that J
m
is nonzero only for odd m
In the case of a pair of wires of small diameter, the expansion (20) has contributions from
all odd integers m. We will make a simplifying assumption that only the term m = 1 is
important, which corresponds to replacing the helical wires by a pair of helical wire bundles,
each of which extends over = , such that the current in the bundles at xed z varies as
cos . If the peak current in each wire is I, then
J(, z, t) =
I
2a cos
e
i
e
i(k
1
zt)
, (21)
E
z
(r < a) = E
1
I
1
(k

1
r)
I
1
(k

1
a)
e
i
e
i(k
1
zt)
, E
z
(r > a) = E
1
K
1
(k

1
r)
K
1
(k

1
a)
e
i
e
i(k
1
zt)
, (22)
and
B
z
(r < a) = B
1
I
1
(k

1
r)
I

1
(k

1
a)
e
i
e
i(k
1
zt)
, B
z
(r > a) = B
1
K
1
(k

1
r)
K

1
(k

1
a)
e
i
e
i(k
1
zt)
. (23)
To deduce the other eld components from the forms (17)-(18) it is useful to note that
the electromagnetic elds can also be derived from from electric and magnetic Hertz vectors
Z
E
and Z
M
(also called polarization potentials; see, for example, sec. 1.11 and chap. 6 of [13]),
each of which has only a z-component. These Hertz scalars, which we call Z
E
and Z
M
, obey
4
the scalar Helmholtz equation, (
2
+ k
2
f
)Z
E
, Z
M
= 0, outside the wires. Thus, the Hertz
scalars also have the forms (22)-(23), and we will verify that
Z
E
=
E
z
k

1
2
, Z
M
=
B
z
k

1
2
. (24)
The scalar and vector potentials V and A are related to the Hertz vectors according to
V = Z
E
, A =
1
c
Z
E
t
+Z
M
, (25)
and hence the electric and magnetic elds E and H are given by
E = ( Z
E
)
1
c
2

2
Z
E
t
2

1
c

Z
M
t
, B =
1
c

Z
E
t
+(Z
M
). (26)
The components of the electromagnetic elds in cylindrical coordinates in terms of the Hertz
scalars Z
E
and Z
M
are (see sec. 6.1 of [13] with u
1
= r, u
2
= , h
1
= 1 and h
2
= r),
E
r
=

2
Z
E
rz

1
cr

2
Z
M
t
, (27)
E

=
1
r

2
Z
E
z
+
1
c

2
Z
M
rt
, (28)
E
z
=
1
r
_

r
_
r
Z
E
r
_
+

_
1
r
Z
E

__
, (29)
B
r
=

2
Z
M
rz
+
1
cr

2
Z
E
t
, (30)
B

=
1
r

2
Z
M
z

1
c

2
Z
E
rt
, (31)
B
z
=
1
r
_

r
_
r
Z
M
r
_
+

_
1
r
Z
M

__
. (32)
For what its worth, the elds associated with Z
E
are transverse magnetic (TM), while those
associated with Z
M
are transverse electric (TE).
To use the forms (22)-(23) in eqs. (27)-(32), we note that
I

m
(k

m
r) = k

m
I
m1

mI
m
r
= k

m
I
m+1
+
mI
m
r
,
1
r
d[rI

m
(k

m
r)]
dr
=
_
k

m
2
+
m
2
r
_
I
m
, (33)
K

m
(k

m
r) = k

m
K
m1

mK
m
r
= k

m
K
m+1
+
mK
m
r
,
1
r
d[rK

m
(k

m
r)]
dr
=
_
k

m
2
+
m
2
r
_
K
m
,
(34)
so that for r < a the eld components are
E
r
=
1
k

1
2
_
ik
1
E
1
I

1
(k

1
r)
I
1
(k

1
a)
+
k
f
r
B
1
I
1
(k

1
r)
I

1
(k

1
a)
_
e
i
e
i(k
1
zt)
, (35)
5
E

=
1
k

1
2
_
k
1
r
E
1
I
1
(k

1
r)
I
1
(k

1
a)
ik
f
B
1
I

1
(k

1
r)
I

1
(k

1
a)
_
e
i
e
i(k
1
zt)
, (36)
E
z
= k

1
2
Z
E
= E
1
I
1
(k

1
r)
I
1
(k

1
a)
e
i
e
i(k
1
zt)
, (37)
B
r
=
1
k

1
2
_
k
f
r
E
1
I
1
(k

1
r)
I
1
(k

1
a)
ik
1
B
1
I

1
(k

1
r)
I

1
(k

1
a)
_
e
i
e
i(k
1
zt)
, (38)
B

=
1
k

1
2
_
ik
f
E
1
I

1
(k

1
r)
I
1
(k

1
a)
+
k
1
r
B
1
I
1
(k

1
r)
I

1
(k

1
a)
_
e
i
e
i(k
1
zt)
, (39)
B
z
= k

1
2
Z
M
= B
1
I
1
(k

1
r)
I

1
(k

1
a)
e
i
e
i(k
1
zt)
, (40)
and for r > a we have the forms (35)-(40) with the substitution I
1
K
1
.
We now see that the continuity of E

and B
r
across the surface r = a, as previously
mentioned, is satised by the above forms.
2.2 Determination of k
0
and the Group and Signal Velocities
The current in the helical windings is assumed to ow only at angle with respect to
the z-axis, so that for good conductors the conductivity of the wires is innite in this
direction, and zero in the perpendicular directions. Hence, the electric eld on the surface
of the cylinder r = a must be perpendicular to the direction of the current, i.e.,
E

(r = a) = cot E
z
(r = a), (41)
and hence,
_
k

1
2
a cot k
1
_
E
1
+ ik
f
aB
1
= 0. (42)
Also, the tangential component of the magnetic eld in the direction of the current must be
continuous at r = a, which implies that
B
z
(r = a

) + tan B

(r = a

) = B
z
(r = a
+
) + tan B

(r = a
+
), (43)
and hence,
ik
f
aI

1
(k

1
a)K

1
(k

1
a)E
1
+
_
k

1
2
a cot k
1
_
I
1
(k

1
a)K
1
(k

1
a)B
1
= 0. (44)
For the simultaneous linear equations (42) and (44) to be consistent, the determinant of the
coecient matrix must vanish, i.e.,
_
k

1
2
a cot k
1
_
2
= (k
f
a)
2
I

1
(k

1
a)K

1
(k

1
a)
I
1
(k

1
a)K
1
(k

1
a)
. (45)
This determines k
0
(and therefore k
1
and k

1
) in terms of a, p and k
f
.
We restrict our attention to low frequencies such that k
f
a 1. In the limit that k
f
and
k
0
vanish, then k
1
= k

1
= k
p
and k
2
p
a cot k
p
= 0, recalling that cot = 1/k
p
a, so that
eq. (45) is satised. For small k
f
and k
0
we approximate
k
1
= k
p
+ k
0
k
p
_
1 +
k
0
k
p
_
, k

1
2
= k
2
1
k
2
f
k
2
p
_
1 + 2
k
0
k
p

k
2
f
k
2
p
_
, (46)
6
so that it suces to take the arguments of the Bessel functions as k
p
a. Using these in eq. (45)
and recalling eqs. (33)-(34), we nd
_
k
0

k
2
f
k
p
_
2
k
2
0
(k
f
a)
2
I

1
(k
p
a)K

1
(k
p
a)
I
1
(k
p
a)K
1
(k
p
a)
= k
2
f
C
2
(k
p
a), (47)
where the constant C dened by
C
2
(k
p
a) = a
2
I

1
(k
p
a)K

1
(k
p
a)
I
1
(k
p
a)K
1
(k
p
a)
=
[k
p
aI
0
(k
p
a) I
1
(k
p
a)][k
p
aK
0
(k
p
a) + K
1
(k
p
a)]
I
1
(k
p
a)K
1
(k
p
a)
(48)
is real and positive since K

1
is negative, as seen in the gure below.
For example, if = 45

then k
p
a = 1, and
C
2
(1)
[1.2 0.55][0.4 + 0.6]
0.55 0.6
2, (49)
and C(1) 1.4.
For k
p
a 1 (gentle twist) then I
0
(k
p
a) 1 + (k
p
a)
2
/2, I
1
(k
p
a) k
p
a/2 + (k
p
a)
3
/8, and
K
1
(k
p
a) k
p
aK
0
(k
p
a), so we have
C
2
(k
p
a 1)
k
p
aI
0
(k
p
a)
I
1
(k
p
a)
1 1 + (k
p
a)
2
/2
1
cos
. (50)
From eq. (47), the wave number k
0
is
k
0
Ck
f
= C

c
. (51)
7
Recalling from eqs. (8)-(9) that k
(1)
k = (k
0
+ k
p
) z

/r, eq. (51) can be recast as the
dispersion relation,
= (k
(1)
) (k)
c
C
k
0
=
c
C
_
k
z

k
p
r
r
_
=
c
C
(k
z
+ k
p
rk

). (52)
Then, the group velocity vector (13) is
6
v
g
=
k
(k) =

k
z
z +

k

c
C
(z + k
p
r

). (53)
While the z-component, v
g,z
of the group velocity is independent of radius r, the group
velocity vector v
g
makes angle
g
to the z-axis given by
tan
g
k
p
r. (54)
At very small r the group velocity is essentially parallel to the z-axis, but at large r lines of
the group velocity form helices with very small pitch. The magnitude of the group velocity
is
v
g

c
C
_
1 + (k
p
r)
2
, (55)
which exceeds c at large r. However, the signal velocity v
s
is clearly
v
s
= v
g,z
=
c
C
< c. (56)
Comparing with eq. (12), we see that the group velocity v
g
is perpendicular to the phase
velocity v
p
, and that on the surface r = a the group velocity is along the direction of the
helical windings.
For = 45

we nd that v
g,z
c/C 0.7c c cos for an uninsulated twisted-
pair transmission line. This happens to be close to the group velocity of typical insulated,
untwisted two-wire transmission lines!
For gently twisted, uninsulated pairs and low frequencies, eqs. (50) and (53) indicate that
v
g,z
c

cos .
2.3 Characteristic Impedance Z
0
at Low Frequencies
To evaluate the characteristic impedance of the transmission line at low frequencies, we
consider the radial electric eld (35) for r < a, for which we need to know the constants B
1
and E
1
in terms of the (peak) current I in the windings.
We can relate B
1
to the (peak) current I in the twisted pair via Amp`eres law for a small
loop of length dz in the r-z plane that surrounds a short segment of the conductor where
the current is maximal:
4
c
I
max, through loop
=
4
c

p
I = |B
z
(r = a

) B
z
(r = a+)| dz
B
1
_
I
1
(k
p
a)
I

1
(k
p
a)

K
1
(k
p
a)
K

1
(k
p
a)
_
dz. (57)
6
The group velocity vector follows straight lines in homogenous media (see, for example, sec. 2.1 of [12]).
Because of the twisted conductors, the present problem is not one of a homegenous medium, and the group
velocity vector eld need not have straight streamlines.
8
That is,
B
1
=
4
c

p
I

1
(k
p
a)K

1
(k
p
a)
I

1
(k
p
a)K
1
(k
p
a) I
1
(k
p
a)K

1
(k
p
a)
I =
4
c
k
p
2a
C
2
DI, (58)
where
D(k
p
a) =
1
a
I
1
(k
p
a)K
1
(k
p
a)
I

1
(k
p
a)K
1
(k
p
a) I
1
(k
p
a)K

1
(k
p
a)
=
I
1
(k
p
a)K
1
(k
p
a)
[k
p
aI
0
(k
p
a) I
1
(k
p
a)]K
1
(k
p
a) + I
1
(k
p
a)[k
p
aK
0
(k
p
a) + K
1
(k
p
a)]
. (59)
Then, eqs. (42) and (51) tell us that
E
1

ik
f
a
k
0
B
1

ia
C
B
1
=
4
c
ik
p
2
CDI. (60)
From eq. (35) we see that the radial electric eld for r < a is largely due to the term in E
1
since k
f
k
1
(at low frequencies). That is,
E
r
(r < a)
i
k
p
E
1
I

1
(k
p
r)
I
1
(k
p
a)
e
i
e
i(k
1
zt)
=
4
c
CDI
2
I

1
(k
p
r)
I
1
(k
p
a)
e
i
e
i(k
1
zt)
. (61)
The peak voltage dierence between the opposing currents is therefore
V = 2
_
a
0
|E
r
| dr
4
c
CDI = Z
0
I, (62)
where
Z
0
377 CD. (63)
When = 45

,
D
0.55 0.6
(1.2 0.44) 0.6 + 0.55 (0.4 + 0.6)
= 0.35, (64)
so that
Z
0
( = 45

) 377 1.4 0.35 = 185 . (65)

In practice, the wires of the twisted pair are insulated, which reduces the characteristic
impedance to 100 .
For gentle twists (k
p
a 1) eq. (59) simplies to
D
I
1
(k
p
a)
k
p
aI
0
(k
p
a)

1
2
, (66)
so that, recalling eq. (50),
Z
0
( 0)
189

cos
, (67)
little dierent from the value at = 45

.
9
2.4 Energy Flux, Momentum and Angular Momentum Density
At low frequencies where k

1
k
1
k
p
k
f
the electromagnetic elds for r < a follow
from eq. (35)-(40) using eqs. (58) and (60) for the constants E
1
and B
1
in terms of the peak
current I,
E
r

4
c
CDI
2
I

1
(k
p
r)
I
1
(k
p
a)
e
i
e
i(kpzt)
, (68)
E

4
c
iCDI
2r
I
1
(k
p
r)
I
1
(k
p
a)
e
i
e
i(kpzt)
, (69)
E
z

4
c
ik
p
CDI
2
I
1
(k
p
r)
I
1
(k
p
a)
e
i
e
i(kpzt)
, (70)
B
r

4
c
iC
2
DI
2a
I

1
(k
p
r)
I

1
(k
p
a)
e
i
e
i(kpzt)
, (71)
B

4
c
C
2
DI
2ar
I
1
(k
p
r)
I

1
(k
p
a)
e
i
e
i(kpzt)
, (72)
B
z

4
c
k
p
C
2
DI
2a
I
1
(k
p
r)
I

1
(k
p
a)
e
i
e
i(kpzt)
, (73)
and for r > a we have the forms (68)-(73) with the substitution I
1
K
1
.
The electric eld components (68)-(70) have similar strength (in Gaussian units) to the
magnetic eld components (71)-(73). The latter correspond to the m = 1 term in the series
expansions for the quasistatic magnetic elds given in [5, 6, 7, 8].
The time-average Poynting vector S for r < a at low frequencies is
S =
c
8
Re(E B

) =
c
8
Re[(E

z
E
z
B

) r + (E
z
B

r
E
r
B

z
)

+ (E
r
B

r
) z]

4
c
C
3
D
2
I
2
4a
I
1
(k
p
r)I

1
(k
p
r)
I
1
(k
p
a)I

1
(k
p
a)
_
k
p

+
z
r
_
, (74)
and that for r > a is obtained from eq. (74) with the substitution I
1
K
1
.
At low frequencies there is no time-average ow of energy in the radial direction, and
hence no radiation is emitted by the transmission line.
7
The energy-ow/Poynting vector (74) is in the same direction as the group velocity (53),
as generally expected.
8
Lines of the Poynting ux S on the cylinder of radius r follow
helices that make angle

g
tan
1
k
p
r (54)
to the z-axis, such that only at r = a does the energy ow in a helix whose angle matches
that of the windings, . At small r the (small) energy ows largely parallel to the axis. At
large r the angle
S
approaches 90

and the Poynting vector is almost entirely transverse;

however because K
1
(k
p
r) 0 at large r there is very little energy associated with these very
tight spirals.
7
Even if we keep the smaller terms in E

and B

of eqs. (36) and (39) there is still no radiation emitted

by the transmission line at low frequencies.
8
See, for example, sec. 2.1 of [12] and references therein.
10
The Poynting vector is at right angles to the wave vector (9), whose angle
k
to the z-axis
is given by eq. (12).
The Poynting vector plays the dual role of describing energy ux and momentum density,
where the latter is given by
p =
S
c
2
(75)
in vacuum. The density l of angular momentum in the electromagnetic eld is therefore
l = r p = r
S
c
2
. (76)
On averaging over azimuth only the z-component of the angular momentum is nonzero,
l =
4
c
C
3
D
2
I
2
4a
I
1
(k
p
r)I

1
(k
p
r)
I
1
(k
p
a)I

1
(k
p
a)
k
p
r
c
2
z. (77)
Thus, the electromagnetic waves on a right-handed twisted-pair transmission line carry pos-
itive angular momentum. In a quantum view, the photons of the wave have angular momen-
tum h and energy h. Hence, we expect that l = (u /) z where u = (|E|
2
+|B|
2
)/8
is the time-average electromagnetic energy density. However, this relation is not self evident
given the description of the waves in terms of Bessel functions.
A Appendix: A Single Wire Helix
We can compare the twisted-pair transmission line to the case of a single helical wire [1, 2]
in the sheath approximation that the helical current ows at angle uniformly over the
entire cylinder r = a, such that the current and elds have no azimuthal dependence. Then,
instead of eqs. (35)-(40) r < a, we now have
E
r
=
ik
1
k

0
2
E
0
I

0
(k

0
r)
I
0
(k

0
a)
e
i(k
0
zt)
, (78)
E

=
ik
f
k

0
2
B
0
I

0
(k

0
r)
I

0
(k

0
a)
e
i(k
0
zt)
, (79)
E
z
= E
0
I
0
(k

0
r)
I
0
(k

0
a)
e
i(k
0
zt)
, (80)
B
r
=
ik
1
k

1
2
B
0
I

0
(k

0
r)
I

0
(k

0
a)
e
i(k
0
zt)
, (81)
B

=
ik
f
k

1
2
E
0
I

0
(k

0
r)
I
0
(k

0
a)
e
i(k
0
zt)
, (82)
B
z
= B
0
I
0
(k

0
r)
I

0
(k

0
a)
e
i(k
0
zt)
, (83)
and for r > a we have the forms (78)-(83) with the substitution I
0
K
0
.
The condition (41) now implies that
k

0
2
E
0
+ ik
f
cot B
0
= 0. (84)
11
Similarly, the condition (43) implies that
ik
f
cot I

0
(k

0
a)K

0
(k

0
a)E
0
+ k

0
2
I
0
(k

0
a)K
0
(k

0
a)B
0
= 0. (85)
The vanishing of the determinant of the coecient matrix tells us that
k

0
4
= k
2
f
cot
2

0
(k

0
a)K

0
(k

0
a)
I
0
(k

0
a)K
0
(k

0
a)
= k

0
2
k
2
f
cot
2

I
1
(k

0
a)K
1
(k

0
a)
I
0
(k

0
a)K
0
(k

0
a)
, (86)
recalling eqs. (33)-(34). That is,
k

_
I
0
(k

0
a)K
0
(k

0
a)
I
1
(k

0
a)K
1
(k

0
a)
= k
f
cot . (87)
At low frequencies such that k
f
a 1 the factor involving Bessel functions in eq. (87) becomes
large, and k

0
k
f
, as illustrated in the gure below, from [1].
Then, k
0
=
_
k
2
f
+ k

0
2
k
f
so that the phase velocity and group velocity are both very
close to c.
References
[1] J.R. Pierce, Theory of the Beam-Type Traveling-Wave Tube, Proc. IRE 35, 111 (1947),
http://puhep1.princeton.edu/~mcdonald/examples/EM/pierce_pire_35_111_47.pdf
[2] L.J. Chu and J.D. Jackson, Field Theory of Traveling-Wave Tubes, Proc. IRE 36, 853
(1948), http://puhep1.princeton.edu/~mcdonald/examples/EM/chu_pire_36_853_48.pdf
[3] S. Sensiper, Electromagnetic Wave Propagation on Helical Structures, Proc. I.R.E. 43,
149 (1955),
http://puhep1.princeton.edu/~mcdonald/examples/EM/sensiper_pire_43_149_55.pdf
[4] M. Chodorow and E.L. Chu, Cross-Wound Twin Helices for Traveling-Wave Tubes, J.
Appl. Phys. 26, 33 (1955),
http://puhep1.princeton.edu/~mcdonald/examples/EM/chodorow_jap_26_33_55.pdf
[5] H. Buchholtz, Electrische und magnetische Potentialfelder (Springer-Verlag, Berlin,
1957), p. 292.
12
[6] A.Y. Alksne, Magnetic Fields Near Twisted Wires, IEEE Trans. Space Elec. Tel. 11,
154 1964),
http://puhep1.princeton.edu/~mcdonald/examples/EM/alksne_ieeetset_11_154_64.pdf
[7] J.R. Moser and R.F. Spencer, Jr, Predicting the Magnetic Fields from a Twister-Pair
Cable, IEEE Trans. Elec. Compat. 10, 324 (1968),
http://puhep1.princeton.edu/~mcdonald/examples/EM/moser_ieeetec_10_324_68.pdf
[8] S. Shenfeld, Magnetic Fields of Twisted-Wire Pairs, IEEE Trans. Elec. Compat. 11,
164 (1969),
http://puhep1.princeton.edu/~mcdonald/examples/EM/shenfeld_ieeetec_11_164_69.pdf
[9] D.F. Alferov et al., The Ondulator as a Source of Electromagnetic Radiation, Part.
Accel. 9, 223 (1979),
http://puhep1.princeton.edu/~mcdonald/examples/accel/alferov_pa_9_223_79.pdf
[10] K.T. McDonald, Bessel Beams (Jan. 17, 2000),
http://puhep1.princeton.edu/~mcdonald/examples/bessel.pdf
[11] R. Warnecke and P. Guenard, Some Recent Work in France on New Typres of Valves
for the Highest Radio Frequencies, Proc. IEE 100, 351 (1953),
http://puhep1.princeton.edu/~mcdonald/examples/EM/warnecke_piee_100_351_53.pdf
R. Warnecke et al., The M-Type Carcinotron Tube, Proc. IRE 43, 413 (1955),
http://puhep1.princeton.edu/~mcdonald/examples/EM/warnecke_pire_43_413_55.pdf
[12] K.T. McDonald, Flow of Energy from a Localized Source in a Uniform Anisotropic
Medium (Dec. 8, 2007), http://puhep1.princeton.edu/~mcdonald/examples/biaxial.pdf
[13] J.A. Stratton, Electromagnetic Theory (McGraw-Hill, New York, 1941).
13